CN-122016086-A - Expressway pavement temperature inversion method and system integrating meteorological information and traffic information

CN122016086ACN 122016086 ACN122016086 ACN 122016086ACN-122016086-A

Abstract

The invention discloses a highway pavement temperature inversion method and system integrating weather and traffic information, wherein the method comprises the steps of obtaining weather information and traffic information of a target highway section, performing time synchronization, space matching and preprocessing on multi-source data, and constructing an inversion input data set; the method comprises the steps of respectively extracting characteristics of meteorological information, road category information and traffic time sequence information based on a multi-input branch neural network, comprehensively processing at a fusion layer, establishing a road surface temperature inversion model, introducing a physical loss function embedded with energy balance constraint in a model training process to enhance physical consistency of inversion results, and carrying out inversion calculation on the road surface temperature of a target road section at corresponding time by using the trained model. The method can comprehensively consider the difference between meteorological conditions and road running environments, improves the accuracy and stability of road surface temperature inversion, and is suitable for monitoring the road surface temperature of the expressway and guaranteeing traffic safety.

Inventors

ZHU SHOUPENG
WU HONG
LIU DUANYANG
JI YAN
ZHU CHENGYING
LV YANG
ZU FAN
JI LUYING
WANG HONGBIN
ZHOU LINYI
XU JINGYUAN

Assignees

南京气象科技创新研究院

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. The highway pavement temperature inversion method integrating meteorological and traffic information is characterized by comprising the following steps of: Acquiring meteorological information corresponding to a target expressway section, wherein the meteorological information comprises air temperature, humidity, wind speed, net radiation and precipitation information; Acquiring traffic information corresponding to a target expressway road section, wherein the traffic information comprises road category information representing road structural characteristics and traffic time sequence information representing traffic operation rules; Performing time synchronization, space matching and quality control on meteorological information and traffic information, performing normalization preprocessing, and constructing a unified inversion input data set for model training; Constructing a multi-input branch neural network-based expressway road surface temperature inversion model facing different data types, designing a physical loss function embedded with energy balance constraint, fusing preprocessed meteorological information with traffic information based on an inversion input data set by utilizing deep learning, and training the inversion model, wherein the multi-input branch neural network structure comprises meteorological information branches, road class information branches, traffic time sequence information branches, a characteristic fusion layer and an output layer, the meteorological information is subjected to nonlinear characteristic extraction through a feedforward neural network and is used for describing complex nonlinear relations between meteorological elements and road surface temperature, the road class information is converted into continuous characteristic vectors through embedding mapping and is used for expressing hidden differences of different road structures on a thermal environment, the traffic time sequence information is processed through the time sequence neural network and is used for extracting operation characteristics evolving along with time, and all branch outputs are combined in the characteristic fusion layer to form comprehensive characteristic representation and obtain an inversion result of the road surface temperature through the output layer; And (3) performing inversion calculation on the road surface temperature according to the weather and traffic information aiming at the target time and the target expressway road section by using the trained expressway road surface temperature inversion model to obtain a road surface temperature inversion estimation result of the corresponding time and road section.
2. The method of claim 1, wherein the meteorological element input vector is constructed based on meteorological information, expressed as: ; Wherein, the And The road segment index and the time index are respectively, A vector is input for the meteorological element at the index, 、、、 And The air temperature, humidity, wind speed, net radiation and precipitation at the index, respectively.
3. The method of claim 1, wherein the road class information is represented using discrete variables as: ; Wherein, the For the indexing of the road segments, For the road class label at the index, The total number of the road categories; the traffic timing information is constructed as a time feature vector expressed as: ; Wherein, the For the time index of the time index, For the traffic timing feature vector at the index, For the hour to which the index corresponds, For the date type code to which the index corresponds, And Representing sine and cosine functions, respectively.
4. The method of claim 1, wherein the time synchronization, space matching and quality control of the weather information and the traffic information are performed, and the normalization preprocessing is performed to construct a unified inversion input data set, comprising: Uniformly mapping data with different time resolutions onto the same time axis through time synchronization processing; The meteorological information is in one-to-one correspondence with the corresponding expressway road sections through space matching; removing abnormal values through quality control; carrying out normalization processing on continuous input variables to enable the variables with different dimensions and value ranges to be in the same scale; After the processing, the weather information, the road category information, the traffic time sequence information and the corresponding road surface temperature observation values are combined to form a unified inversion input data set.
5. The method of claim 1, wherein the weather information is processed through a feed-forward neural network to extract its effective features, expressed as: ; Wherein, the A feature vector representing the branch output of the weather information, A feed-forward neural network representing a branch of weather information, For the input of the weather information, A parameter set which is a meteorological information branch; the road category information is mapped through the embedded layer, discrete category information is converted into continuous vector representation, and the form is as follows: ; Wherein, the Is an embedded vector corresponding to the road class information, Is an embedded matrix which is embedded in the matrix, Is road category information Corresponding one-time thermal coding; The traffic time sequence information is processed through the long-term and short-term memory network and is used for capturing the evolution characteristics of traffic change along with time, and the evolution characteristics are expressed as follows: ; Wherein, the Is the characteristic vector of the traffic time sequence information branch output, Is a time feature extraction function of a long-short-period memory network, Is from the moment To the point of Is provided with a traffic timing information input, Is a parameter set of the timing branch; The output characteristics of all branch networks are fused in a splicing or weighting summation mode to obtain a comprehensive characteristic vector : ; Fused feature vectors Input to an output layer, and finally output the road surface temperature inversion estimation : ; Wherein, the Is a mapping function of the output layer, Is the parameter set of the output layer.
6. The method of claim 1, wherein the physical loss function is embedded with an energy balance constraint Expressed as: ; Wherein, the Fitting loss to data, representing road surface temperature inversion estimation With measured road surface temperature Mean square error within a training batch; for the weight coefficient of the physical constraint, For physical constraint loss, represent All sample absolute values within a training batch are averaged; As an energy balance residual, expressed as: ; Wherein, the In order to be a net radiation of the radiation, In order to sense the heat flux, Is geothermal flux; ; ; Wherein, the In order to achieve an air density of the air, The specific heat is fixed for the air pressure, For the inversion estimation of the road surface temperature, And The road segment index and the time index are respectively, For the air temperature at the index, Is a turbulent thermal resistance, Is the equivalent parameter of the heat conductivity coefficient, Is the temperature of the pavement under layer.
7. Highway road surface temperature inversion system of fusion meteorological and traffic information, its characterized in that includes: the weather information acquisition unit is used for acquiring weather information corresponding to the target expressway road section, wherein the weather information comprises air temperature, humidity, wind speed, net radiation and precipitation information; The traffic information acquisition unit is used for acquiring traffic information corresponding to the target expressway road section, wherein the traffic information comprises road category information representing road structural characteristics and traffic time sequence information representing traffic operation rules; The data set construction unit is used for carrying out time synchronization, space matching and quality control on the meteorological information and the traffic information, carrying out normalization preprocessing, constructing a unified inversion input data set and being used for model training; the system comprises a model construction and training unit, a data processing unit and a data processing unit, wherein the model construction and training unit is used for constructing expressway road surface temperature inversion models based on multi-input branch neural networks for different data types, designing physical loss functions embedded with energy balance constraint, utilizing deep learning to fuse preprocessed meteorological information with traffic information and train the inversion models, the multi-input branch neural network structure comprises meteorological information branches, road class information branches, traffic time sequence information branches, a characteristic fusion layer and an output layer, the meteorological information is subjected to nonlinear characteristic extraction through a feedforward neural network and is used for describing complex nonlinear relations between meteorological elements and road surface temperatures, the road class information is converted into continuous characteristic vectors through embedding mapping and is used for expressing implicit differences of different road structures on a thermal environment, the traffic time sequence information is processed through the time sequence neural network and is used for extracting operation characteristics evolving along with time, the branch outputs are combined in the characteristic fusion layer to form comprehensive characteristic representation, and an inversion result of the road surface temperatures is obtained through the output layer; The inversion calculation unit is used for carrying out inversion calculation on the road surface temperature based on weather and traffic information aiming at the target time and the target highway section by using the trained highway road surface temperature inversion model to obtain the road surface temperature inversion estimation result of the corresponding time and the corresponding section.
8. An electronic device, comprising: A memory for storing a computer program; a processor for executing the computer program to implement the method of any of claims 1-7.
9. A non-volatile storage medium for storing a computer program, wherein the computer program when executed by a processor implements the method of any of claims 1-7.
10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the method of any of claims 1-7.

Description

Expressway pavement temperature inversion method and system integrating meteorological information and traffic information Technical Field The invention relates to a highway pavement temperature inversion method, in particular to a highway pavement temperature inversion method and system integrating weather and traffic information. Background The road surface temperature is an important environmental factor affecting the traffic safety of the expressway, and especially under the conditions of low temperature, rain and snow and day and night alternation, the tiny change of the road surface temperature can directly lead to the icing, snow accumulation or wet skid of the road surface, thereby causing traffic accidents. At present, the expressway road surface temperature acquisition mainly depends on a point-type road surface temperature sensor, or a simple model inversion method which only considers a small amount of meteorological elements such as air temperature, radiation and the like, and has limited inversion precision under complex weather conditions and diversified road structure conditions. In fact, radiation environment, ventilation condition, road surface heat exchange characteristic, traffic flow and the like can directly or indirectly affect the road surface heat environment, and the prior art is lack of a road surface temperature inversion method for simultaneously fusing weather information and traffic information. Disclosure of Invention The invention aims to provide a highway pavement temperature inversion method and system integrating weather and traffic information, which are used for realizing the refinement and dynamic inversion of the highway pavement temperature by comprehensively considering weather elements and traffic information and establishing a pavement temperature inversion model. The technical scheme is that the method comprises the following steps: Acquiring meteorological information corresponding to a target expressway section, wherein the meteorological information comprises air temperature, humidity, wind speed, net radiation and precipitation information; Acquiring traffic information corresponding to a target expressway road section, wherein the traffic information comprises road category information representing road structural characteristics and traffic time sequence information representing traffic operation rules; Performing time synchronization, space matching and quality control on meteorological information and traffic information, performing normalization preprocessing, and constructing a unified inversion input data set for model training; Constructing a multi-input branch neural network-based expressway road surface temperature inversion model facing different data types, designing a physical loss function embedded with energy balance constraint, fusing preprocessed meteorological information with traffic information based on an inversion input data set by utilizing deep learning, and training the inversion model, wherein the multi-input branch neural network structure comprises meteorological information branches, road class information branches, traffic time sequence information branches, a characteristic fusion layer and an output layer, the meteorological information is subjected to nonlinear characteristic extraction through a feedforward neural network and is used for describing complex nonlinear relations between meteorological elements and road surface temperature, the road class information is converted into continuous characteristic vectors through embedding mapping and is used for expressing hidden differences of different road structures on a thermal environment, the traffic time sequence information is processed through the time sequence neural network and is used for extracting operation characteristics evolving along with time, and all branch outputs are combined in the characteristic fusion layer to form comprehensive characteristic representation and obtain an inversion result of the road surface temperature through the output layer; And (3) performing inversion calculation on the road surface temperature according to the weather and traffic information aiming at the target time and the target expressway road section by using the trained expressway road surface temperature inversion model to obtain a road surface temperature inversion estimation result of the corresponding time and road section. Further, a meteorological element input vector is constructed based on meteorological information, expressed as: ; Wherein, the AndThe road segment index and the time index are respectively,A vector is input for the meteorological element at the index,、、、AndThe air temperature, humidity, wind speed, net radiation and precipitation at the index, respectively. Further, the road class information is expressed as using discrete variables: ; Wherein, the For the indexing of the road segments,For the road class label at the index,The total number of the road